Chapter 16

Coordinating Health Services: An Operations

Management Perspective
Thomas R. Rohleder, David Cooke, Paul Rogers, and Jason Egginton

1 Introduction
The rising costs of healthcare in the USA and around the world are well
documented. A recent report from the Commonwealth Fund (Davis et al. 2010)
showed that costs throughout the world have been rapidly increasing in recent
decades. Driving these cost increases are new technologies, tests, patient access,
inefficiencies, and myriad other factors. Such cost increases might be justified if
the healthcare delivered was of higher quality. However, this is highly
questionable. For example, in the previously referenced report by Davis et al.,
the USA was last of seven major developed countries in overall health system
performance (see Fig. 16.1). Yet, the cost of healthcare in the USA is nearly
double the cost of the top performer, the Netherlands.
The Commonwealth Fund Report has several different categories that make up
health system performance. One of the components of the Quality Care category is
Coordinated Care. Based on the rankings in Fig. 16.1, it appears that this
component may have some influence on both the quality of care and healthcare
expenditures. In the context of the Commonwealth Report, Coordinated Care
focuses on how well care is managed among health services from a medical
perspective. However, from an operations research perspective, coordination is
also about having the right

T.R. Rohleder ([81) J. Egginton

Division of Healthcare Policy and Research, Department of Health Sciences Research,
Mayo Clinic, Rochester, MN, USA
e-mail: Rohleder@mayo.edu
D. Cooke
Cooke Research & Consulting Inc., Calgary, AB, Canada
P. Rogers
Department of Mechanical Engineering, University of Calgary, Calgary, AB, Canada
B.T. Denton (ed.), Handbook of Healthcare Operations Management: Methods
421
and Applications, International Series in Operations Research & Management Science 184,
DOI 10.1007/978-1-4614-5885-2 16, Springer Science+Business Media New York 2013

42
2

T.R. Rohleder et al.

Fig. 16.1 Report on healthcare system performance: Commonwealth Fund (2010)

quantities of healthcare resources and how well they are managed together. This
latter perspective of coordination affects the former by ensuring patients have
access to the needed healthcare resources and flow through them effectively and
efficiently.
Why is this type of coordination important? There are many reasons, including
avoiding duplication of services. Without coordination, the same information from
patients may be gathered at several points, tests may be repeated, and redundant
health services may be provided, resulting in higher cost and burden to patients.
In addition, a lack of coordination may lead to delays in patient treatment. Chronic
shortages of key resources may lead to bottlenecks and long waiting times. As a
way to avoid waiting, patients may use (or be referred to) inappropriate and more
costly resources. For example, shortages of primary care capacity often lead to
increased use of emergency departments and similar more costly and inappropriate
resources (Cheung et al. 2011). Thus, poorly coordinated healthcare resources
may lead to poor quality and higher costs.
The various symptoms of a lack of coordination may lead to lower satisfaction
for patients and result in less than optimal treatment. Potentially, deadly errors
could occur if patients receive conflicting medications or treatments. Poor
coordination is one of the issues discussed in the report To Err is Human:
Building a Safer Health System (IOM 1999) that notes One oft-cited problem
arises from the decentral- ized and fragmented nature of the healthcare delivery
systemor nonsystem, to some observers.

Lack of coordination of healthcare resources has been considered from many

perspectives. Clinicians look at the issue from the perspective of the frontline
healthcare provider and consider improved communication and information
sharing as a key to improvement (Sucliffe et al. 2004). Related to better
communication is the push for better patient record sharing and management,
particularly better electronic medical records (Hillestad, et al. 2005). However,
we propose to look at coordination from an operations management (OM)
perspective and focus on methods that simultaneously increase effectiveness and
reduce costs.
In this chapter we will focus on how operations research can improve service
coordination in health systems. An overview is provided to show how operations
research contributes to value in healthcare systems. Next, we highlight previous
research that looks specifically at health services. Two case studies will focus on
different aspects of coordination:

Case one uses system dynamics modeling to show the importance of

coordinating the right capacity levels of resources and the need for long-term
planning models in a regional health system.
Case two uses discrete-event simulation to show the value of coordinating
health services within a hospital to ensure timely patient access.
Finally, we will discuss future challenges in coordinating health services that
are being driven by changes in healthcare delivery models such as the medical
home, shared decision making, and centralization of health services.

2 Background on Health Systems

Operations management within healthcare can be viewed from the traditional
trans- formation process perspective (Meredith and Shafer 2007), as shown in Fig.
16.2. Various staff, facilities, technology, and patients are inputs to the healthcare
trans- formation process. The healthcare delivery process is the value-adding
stage and includes how the inputs are managed. The effectiveness of delivery
strategies and policies and how these are executed in the healthcare operations are
key in creating valuable outputs. Treating patients, medical and healthcare service
research, and training of healthcare staff are all common outputs for healthcare
organizations. In the end, value is defined by the patient and healthcare market.
Effective coordination of the inputs is a key to achieving high-value healthcare.
High value can be defined as achieving high system performance and low per
capita costs (as highlighted in Fig. 16.1).
A complication with understanding value in healthcare is that it entails multiple
dimensions. Certainly improved patient health or condition is one dimension of
value. However, patient perceptions of what treatment was received and how this
treatment was delivered are another dimension and may be at odds with the health
outcome. An arduous treatment process or treatment that does not use the inputs
desired by the patient may cause dissatisfaction even though the health outcome

INPUTS

VALUE ADDING

OUTPUTS

Human
Resources

Knowledge/
Skills

Equipment/
Technology

Capital ($)

TRANSFORM INPUTS
TO OUTPUTS
Administration's role:
-

Treated Patients
Research
Training

VALUE

System coordination
Set direction
Policies
Control/Measurement

Materials/
Supplies
FEEDBACK
Information

Fig. 16.2 The operations management transformation process

is positive. In this chapter we will consider value in a multidimensional context

that includes service quality, timeliness, price/cost, as well as treatment outcomes.
However, we will assume that patient outcomes are highly correlated with patient
satisfaction even though for individual patients this may not always be true.
Tension between the individual versus population-based orientation in
healthcare is another complication to understanding value. This is particularly
relevant when a health system has significant resource constraints. If the capacity
for a technology like positron emission tomography (PET) is scarce, then it
makes sense from a population perspective to allow only the most appropriate
patients access to this diagnostic/treatment. This may exclude some patients
from receiving the treatment, even though they could benefit. Thus, individual
value may be sacrificed to maximize population or system-wide value. We will
consider value from both perspectives, but focus on the system or population
perspective.

2.1 Literature Review

In Building a Better Delivery System: A New Engineering/Healthcare Partnership, the Institute of Medicine and the National Academy of Engineering (2005)
partnered with the goal . . . to transform the U.S. healthcare sector from an

underperforming conglomerate of independent entities (individual practitioners,

small group practices, clinics, hospitals, pharmacies, community health centers
et. al.) into a high performance system in which every participating unit
recognizes its dependence and influence on every other unit. The report notes the
importance of systems engineering, operations research, and operations
management as a means to achieve health systems coordination at both the interand intraorganizational levels. The report is US based; however, we believe the
basic philosophy is globally relevant. To emphasize the IOM/NAEs suggested
direction, this review will focus on operations research-oriented literature that
emphasizes resource coordination.
2.1.1 Macrolevel Health Systems Modeling
In this section we discuss macrolevel modeling that considers flows and decision
making across organizations and/or major health services functions. The focus is
on policy level decisions such as overall capacities of the organizations and
functions (e.g., hospitals, primary care). The following section will focus on
microlevel modeling that emphasizes flows within organizations or patient
pathways. To some extent, this delineation is artificial, and there will be
overlaps between problem domains and the tools used for analysis. However,
there is likely affinity for certain methods in each modeling level, and therefore,
we believe separating them provides a useful structure.
Considering the integration of healthcare systems requires conceptual and
quantitative methods that incorporate their inherent breadth and complexity. While
detailed operations problems in healthcare can rely on available or collectable data,
often healthcare systems entail elements where data are not available. Thus,
research and planning methods need the capability to include relationships
without data support.
System dynamics (SD) is a methodology that meets these requirements. Its
simple set of constructs made up (primarily) of stocks and flows are designed for
high-level model building (Sterman 2000). Stocks are accumulations in systems
and in healthcare can represent tangible elements such as patients waiting for
treatment or intangible elements such as physicians commitment to safety. The
flows control the dynamics of the changes in the stocks by specifying the rate
of transition from one stock to another. In healthcare, systems flows can represent
service rates controlled by the capacity of resources such as physicians or
diagnostic equipment. Flows could also be the motivational forces that cause an
increase or decrease of intangible stocks such as commitment to safety. As such,
the tool is effective for considering policies that address both healthcare system
structure and infrastructure. Further, the method is designed to incorporate the
effects of time lags and feedback (Morecroft 2007). All of these aspects are
prevalent when considering a systems perspective of healthcare.

An area where the SD approach has been applied to healthcare systems is

coordinating urgent and elective (or scheduled) care. Brailsford et al. (2004) used
an SD model focusing on the accident and emergency department (ED) of a
hospital in England. The model showed that if growth in ED patients continued,
it would require the eventual reduction of elective admissions to the hospital and
missing government prescribed patient access performance targets. The model was
also used to evaluate scenarios where some patients were seen in community-based
diagnostic centers rather than the hospital ED. Thus, this example shows some of
the strengths of SD modeling: analyzing the changes in systems over an extended
period of time and exploring broad-based policy options.
Lane et al. (2000) used SD to consider a similar environment to that of
Brailsford et al.; however, their model focused on the system effects of hospital
bed reductions on the ED. The authors noted that because of patient priority
policies, decreasing inpatient beds did not have the expected effect of delaying ED
patients; rather it caused elective surgery cancellations. Thus, an important,
generalizable conclusion from this work is that healthcare systems should be
considered from a holistic perspective that assesses performance of all system
components.
The conceptual system model developed by Cooke et al. (2010) considered
a similar environment to those of Brailsford et al. and Lane et al.; however, it
emphasized the role of changing patient demographics in explaining healthcare
system capacity problems. An aging population placed increasing demands on the
EDs in the health system studied. Poor planning and coordination of the capacities
of primary care, hospital beds, and medical specialties created surprising delays in
the EDs given the new demands on the healthcare system. Using causal loop
diagrams that showed the complex interrelationships and feedback effects among
the health system elements and supporting health system data, the Cooke article
hypothesized the many reinforcing factors that created waiting time and access
issues at EDs in a Canadian health system. A more complete quantitative model
developed for this environment will be discussed in Sect. 3.
Wolstenholme (1993) developed a similar model with a focus on how elderly
patients flowed in a healthcare system in the UK. This study showed the value
of system dynamics modeling tools to enhance systems thinking. The model
identified that as posthospital care of elderly in nursing homes was passed on
to budget constrained community care entities, the indirect effect would be to
transfer even higher costs back to the National Health System (NHS) via a need
for more hospital beds and associated resources. This article, again, reinforces the
theme of coordinating multiple elements of systems for achieving effective health
systems.
2.1.2 Microlevel Health Systems Modeling
System dynamics is often a good choice of methodology for the macrolevel and
where its simple constructs have sufficient explanatory capability. However, more
detailed modeling methods such as queuing analysis and discrete-event simulation

(DES) are often required for studying integration of healthcare services within
organizations, healthcare functions, and patient pathways. Brailsford et al. (2004)
used DES to evaluate the efficacy of streaming patients with minor issues to
separate resources within EDs. This model addressed the ED as a system within a
hospital that cares for patients with significantly different levels of acuity.
Streaming these different patient types is increasingly considered a practical way
to improve performance (Kelly et al. 2007). This type of detailed policy is
generally not easy or appropriately modeled using SD.
Levin et al. (2011) presented a case study in which they considered the effect
of increasing heart surgery volumes on ED patient access to cardiology services.
They used regression analysis to model the relationship of patient characteristics
and length of stay (LOS) at various care resources. This analysis was incorporated
into a DES model that explored options of increasing capacity and reducing patient
LOS on patient access. A unique aspect of this study is that it considered how
the resources and operations issues in one department (surgery) affect the service
received in another (ED).
Another way that detailed modeling has been applied to healthcare coordination
is looking at how upstream decisions affect downstream performance. The article
by Bekker and de Bruin (2010) used queuing analysis to report on a number of
analysis scenarios, including how operating room scheduling influences the
number of ward beds required for recovering patients. Using typical operating
room schedules (i.e., no elective surgeries on weekends), the authors show how
changing these schedules affects the number of ward beds required. In particular,
they show that front loading surgeries on Mondays and Tuesdays lead to a more
balanced load on the wards and less overall beds are required. Haraden and Resar
(2004) discussed the various issues integrating hospital services to improve
patient flow, in particular the need for better management of variability in
volumes that are created by hospital staff and decision makers. One example from
the article noted that in most hospitals, elective surgery schedules are based on
individual preferences and do not consider the downstream patient demands in the
ICU and other recovery resources. Without coordination, the downstream recovery
resources simply have to react to the highly variable upstream decisions.
Rohleder et al. (2007) discussed facility design and implementation decisions
associated with building and locating patient service centers (PSCs) for laboratory
testing. Opening new service centers in a Canadian city where existing service
centers were being phased out led to initial underutilization and very high initial
patient satisfaction at the new PSCs. However, the dynamics of patient visitation
eventually led to lower satisfaction and patient complaints. By considering the
implementation and facility design decisions together rather than sequentially,
some of the issue could have been anticipated and the dissatisfaction mitigated.
Assisting decision making for this kind of coordination may also require several
methodologies. Like Brailsford et al. (2004), this study used both SD and DES
methodologies.

3 Case Studies
This section will discuss two case settings showing the value integration of health
services and the potential of operations management concepts and techniques to
assist decision makers. The first case will discuss the use of system modeling
to diagnose and provide policy decision support for a large Canadian urban
healthcare system. The second case will look at coordination of patient pathways
for downstream planning of patient services.

3.1 A System Dynamics Model of the Calgary Health Region

Emergency Departments
Rohleder et al. (2009) expanded on the qualitative model described in Cooke et al.
(2010) to create a macrolevel system dynamics model of the Calgary Health
Region (CHR). The original purpose of the model was to explain the causes for
delays in treating patients in the citys EDs; however, as the study progressed, it
was evident that the scope needed to consider the whole of the major health
services in the region. Therefore, this case shows the importance of looking at
healthcare from a systems perspective and the importance of coordinating
capacity levels across all health services.
At the start of the study, the average LOS of patients in the largest ED was 8 h,
and the average time for initial treatment by an ED doctor was about 1.5 h. Due to
the effects of variability, sometimes the wait time before being seen by a physician
was even longer and led to some unfortunate outcomes in the ED waiting rooms
(Lang 2006). An original hypothesis of the causes for delays is shown in Fig. 16.3.
An aging demographic was believed to be driving up the overall severity level
of patients in Calgary EDs. The proportion of patients that had higher acuity and
greater complexity increased. This increased the load on the ED resources and
therefore the number of patients in the waiting rooms.
Initial data on patient ages and acuity were collected, and they supported the
model in Fig. 16.3; however, the full scope of the problem quickly expanded to
include components of the healthcare system outside of the ED. As the model
evolved, the focus remained on the congestion in the EDs. It also showed the
interconnectedness within health systems and the need to plan and implement
integrated services.
While decision makers continued to try and resolve the problems by improving
ED operations, it became clear from our model that over time the ED became the
safety net to accommodate the demographic effect on healthcare capacity from a
growing population of older and sicker patients. The model helped explain why the
EDs patient access performance was declining rapidly when year-over-year
patient volume growth was almost flat. Essentially, the performance problems in
the EDs were a symptom of capacity and coordination problems throughout the
healthcare system.

Patient Complexity
(Older)

Multiple
Consults

Patient Acuity
(Sicker)
Proportion of ED
Visitors that are
Old and Sick

Older/Sicker
Patients Take
Longer

Leave Without
Being Seen

ED Workup
Time

Consult Workup
Time

ED Length
of Stay

B1
Waiting Room
Overcrowding

Fig. 16.3 Causal loop diagram to explain delays in EDs

The purpose of the system model was to examine the impact on the EDs of
patient flows in the overall CHR system. For example, insufficient inpatient
capacity for elective surgeries increased the possibility that patients on the hospital
waiting list had to go to an ED. Similarly, if a patient was unable to access a family
physician, then they were more likely to go to an ED or urgent care center. In
this study we were interested in long-term impacts of changes in population
demographics and available health services. Thus, we were not concerned with
the hour-to-hour fluctuations in demand for health services, but rather with the
month-to-month and year-to-year ability of the system to cope with the average
demand for services from the regional patient population. The following
subsections will briefly discuss the model building process and results from the
model related to coordination and operations management.
The qualitative model that served as the basis for the SD model was developed
using an iterative group process that required many interviews with people
involved in the healthcare system and bringing groups of healthcare staff and
administrators together to understand the patient flows and factors influencing
patient access. Figure 16.4 shows a highly simplified version of the model of
the patient flows that evolved from this process and became the basis for the
quantitative SD model. Three sub-models emerged from the overall systems view.
The ED care sub-model is the set of stocks and flows along the bottom of Fig.
16.4; the acute care sub-model included the stocks and flows associated with
hospital services (right side of figure); and the primary care sub-model includes
the primary care, urgent care (and other clinics), and specialist consulting services
(top left of figure). Splitting the model

43
0
Discharge to Outpatient
Clinics for Final Therapy
Patients in Outpatient Clinics or
Seeing Consultants

Consultant
Referrals to ED

Outpatient
Discharge

Becoming
Urgent

Primary Care
Visits

Patients on Hospital
Waiting List

Wait to
Admit
FP Referrals to
Consultants or Clinics

Primary Care
Discharge

OR
Scheduling

Direct

Primary Care
Patients
Clinic or Consultant
Admitting

Urgent Care
Visits

Out of Region
Patients

Patients in
Surgery and
Recovery

Admitting
Surgical

Direct
Admitting

Urgent Care
Patients

Urgent Care Referrals to

Consultants or Clinics

Patients Waiting for Clinic or Consultant

Non Surgical
Post Surgery
Admitting

Urgent Care
Discharge
ED Discharge
without Consult to
Outpatient Care

Urgent Care
Referrals to ED

FP Referrals
to ED

Non Surgical
Admitting

Patients in Hospital Wards

ED Discharge after
Consult to
Outpatient Care

ED
Discharge
To Home

ED Walk In
Patients

Decision t
Consult
Left Against
Medical Advice

Patients in ED Bed
Waiting Rooms and
ED
Waiting for ED MD
EMS Hallway
Admitting

Being Seen

ED Discharge
After Consult
to Home

Patients in ED Being
Assessed or Treated

Patients in ED

First
Assessment

ED Hospital
Admitting

Patients in ED

ED Patients Waiting

Receiving or Waiting
for Consult

for Inpatient
Admission

Decision to
Admit
ED Discharge to
ALC or Home Care

Fig. 16.4 Systems model of the Calgary Health Region (with detail of the emergency department)

Wards Discharge
to ALC

Wards Discharge
to Home

T.
R.
Ro
hle
der
et
al.

Fig. 16.5 Data for emergency physician time by CTAS (15) and age

into sub-models made it easier to handle the complexity of the entire system. The
ED sub-model has greater detail due to the initial focus on this area. Nonetheless,
the important flows from each service sector are included.
With the model components and boundaries established, sources of available
data were identified to populate the model. To incorporate the urgency of a
patients condition, we included patient age and an urgency score based on the
Canadian Triage Assessment Scale (CTAS) as patient attributes. Together these
account for much of patients usage of healthcare resources. The CTAS score of a
patient has a particular meaning in the ED where it is used to prioritize patients;
however, we used CTAS throughout the model to determine when patients
required urgent or acute care.
As an example of how we used data in the model, we will use the time required
for an ED physician to perform a first assessment on a patient. The time spent
waiting for an emergency physician (EP) is called mean EP time. The model uses
EP time data that were collected from an ED during the AprilSeptember 2006
time period, involving treatment of over 31,000 patients. Figure 16.5 shows that
the mean EP time is about 23 min for CTAS 1, about 58 min for CTAS 2, and
about
100 min for CTAS 35 patients. We assume no correlation with age in this
situation; however, for other flows, age was an important factor. Although there
appears to be some correlation with age for CTAS 4 patients, the data is skewed by
just a few very old patients with long wait times. For CTAS 4 patients aged 80
and under, there is little or no correlation between patient age and EP time.
Probabilitydistribution functions were used in the model to represent CTAS 15
patients.
The overall model was built in this manner, incorporating relationships between
age and patient acuity as appropriate. As identified in Fig. 16.4, the stocks in the
system are where patients accumulate and wait for access, and the flows control

ED Admitting

People/Day

720
700
680
660
640

Actual ED admitting in CHR (people/day)

620

Model ED admitting (people/day)

600
2000

2001

2002

2003 2004
Year

2005

2006

2007

Fig. 16.6 Comparison of ED admitting rates for model validation

the movement of patients between stocks. Together, these stock and flow
constructs required hundreds of data elements.
Toensure the model created was valid, we compared model results to actual
results for several measures including ED admission rate, inpatient admission
rate, and ED and inpatient LOS, among others. Figure 16.6 shows the graphical
comparison for the average number of patients admitted to the ED per day. The
model reports slightly better performance than occurred in practice. In part this is
due to the fact that the SD model does not incorporate the effects of short-term
variability where within-day peaks and valleys have a negative impact on resource
utilization due to congestion. Nonetheless, the differences were deemed acceptable
to decision makers.
Having a valid model is not enough to ensure its usefulness. A key concern with
use of the model for making system-wide decisions was, would it do a better job
of explaining system behavior than the simple, standard measures in current use?
Typically, health system planners use metrics, such as number of beds per unit
population, to plan the capacity of the healthcare system. A common expectation
in practice is that if the growth in the number of staffed beds grows with the
population, and support services such as laboratory testing and diagnostic imaging
keep pace with bed growth, then capacity needs should be met. To show why the
SD model may be more useful than such simple rules of thumb, Fig. 16.7 reports a
comparison of various key measures related to population changes and healthcare
system performance.
Bed growth kept pace with general population growth over the 20012006
period, but the ED admitting rate did not. A good measure of system capacity is
throughput, and in the case of hospital EDs, this is measured by the ED admitting
rate per day. While the SD model results are not exact, Fig. 16.7 suggests that they
do a better job of predicting actual ED admitting rates than would be expected
from a bed growth projection. As noted in Fig. 16.7, part of the reason for the
inability of bed growth projections to satisfy future demographic demands is that

Calgary Health Region

Historical Growth
1.3
1.25

Population
ED Beds
(model)

50+ Population
Inpatient Beds
ED Admitting (actual)
ED Admitting

1.2
1.15
1.1
1.05
1
0.95
0.9

2001

2002

2003

2004

2005

2006

Fig. 16.7 ED admitting compared to population and bed growth

Table 16.1 Bed policy scenarios
Scenario number

Scenario name

Brief description

Base 2008

ED expansion only

Hospital expansion only

4
5

ED and hospital expansion

Further hospital expansion

No change in ED or hospital capacity (beds

and staffing) from 2008 to 2016
ED capacity is expanded by 20% to
match 20082016 population growth
Hospital capacity expansion to 2011 per
CHR plan
Combination of cases 2 and 3
As case 4, but further 10% hospital capacity
expansion 20142016

the aging population is growing at a much faster rate than the general population.
The system dynamics model is able to take these demographic changes and
system- wide capacity levels that influence ED patient demand into account.
Our focus in the analysis was on the effect of bed capacity on health system
performance; however, as described later, the model could be used to explore
many facets of resource coordination including staffing. We considered five
scenarios, summarized in Table 16.1, which represent different possible
alternatives for staffed bed expansion. Examination of the results will illustrate
the use of the system dynamics model for policy evaluation.
Scenario 1 provides a baseline projection of what would happen if no capacity
expansion took place beyond what existed at the end of 2007. Scenario 2 assumes
that ED bed capacity is expanded at the same rate as population growth. As shown
in Fig. 16.7, this essentially replicates historical practice. Scenario 3 assumes that
inpatient bed capacity is expanded according to projections provided by seniorlevel decision makers. This projection takes into account the planned hospital
expansions.

Hospital Discharge Rates

Base 2008

ED Expansion Only

ED and Hospital Expansion

Further Hospital Expansion 2014-2016

Hospital Expansion Only

CHR Population
1600000

400

1200000
300
1000000
250
800000
200

Population (People)

Discharge Rate (People per Day)

1400000
350

600000
150
400000
100

50

200000

0
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Year

Fig. 16.8 Hospital discharge rates for bed expansion scenarios

Scenario 4 is a combination of scenarios 2 and 3, and these three cases will let us
answer the question, is it better to expand ED beds, inpatient beds, or expand both
together? Finally, scenario 5 was added because the results of scenario 4 revealed
that capacity limitations will again be encountered in 2014 unless further inpatient
bed capacity is added. Scenario 5 assumes the same ED bed capacity as in cases 2
and 4.
In all scenarios we use the actual CHR population figures for 20012006 and
the CHR population projections for 20082016. We assume that there is sufficient
support staff and services to operate the beds and all five scenarios have the
same assumptions for capacity increases in family physicians, urgent care centers,
consultants, and outpatient clinics.
An important issue is the effect of bed expansion on hospital discharge rates, a
measure of patient throughput. In Fig. 16.8 we compare the hospital discharge
rates for each of the five cases to the growth in population rate. The main
conclusions that can be drawn are as follows:
Without inpatient bed expansion, throughput will actually drop. This is because
in a capacity constrained system, the sickest people will be given priority and
they will consume more resources and stay longer than the less sick people who
are displaced. This point emphasizes the need to look at the healthcare services
as a coordinated system.
Adding ED beds does nothing to change the throughput of hospital inpatients.
The planned 20082011 expansions (inpatient beds) will close the gap that has
developed between population growth and hospital throughput during the 2001
2007 period.

Average Fraction LWBS

0.4
0.3

1
1

0.2

1
1
2

12

0.1

1234
4 55 1
123

5 123451 2
34
2

5
34

2
51

3451

2 34
5

45

45

3
4

3
45

3
3

4
4 5

4 5

4 5

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Year (Year)
Average Fraction LWBS : Base 2008
Average
Average
Average
Average

Fraction
Fraction
Fraction
Fraction

LWBS
LWBS
LWBS
LWBS

:
:
:
:

1
1
1
1
1
1
1
1
1
1
1
Dmnl
ED Expansion Only
2
2
2
2
2
2
2
2
2
2
Dmnl
3
3
3
3
3
3
3
3
3
Hospital Expansion Only
Dmnl
4
4
4
4
4
4
4
4
4 Dmnl
ED and Hospital Expansion
5
5
5
5
5
5
5
5 Dmnl
Further Hospital Expansion 2014-2016

Fig. 16.9 Patients who leave the ED without being seen

While capacity from the 20082011 planned expansions will be sufficient to

support population growth to 2013, further capacity expansion in the 2014
2016 time frame will be required to prevent future system bottlenecks.
Another major concern for ED performance is the proportion of patients who
leave the ED without being seen (LWBS). Figure 16.9 shows the results for this
measure for the five scenarios. The model picks up the degradation in performance
in 2006 when several highly publicized adverse events occurred (including a death
that was partly attributed to a patient who left an ED, Lang 2006). Again, it is
clear that a coordinated planning perspective that considers increasing both ED
and hospital beds together is required. Further, additional expansion to that which
is planned is needed to avoid running into severe access problems in the future.
Our model emphasized the role that bed capacities and their coordination play
in the access of integrated health services. However, we could have easily used
the model to explore a variety of strategic level decisions beyond bed capacities.
For instance, there was a prevailing perception that there was a chronic shortage of
primary care services in the CHR. Our model could explore the effect of
increasing primary care capacity on patient access. The effect of adding any type
of the resources in the model versus any other could help shape recruiting and
future system design decisions of a long-term strategic nature.
Overall, this case helps to show the importance of understanding the systemwide flow of patients when considering a single service. System dynamics is a
good tool for this purpose due to its ability to incorporate the relationships between
many system elements, including feedback from other system elements. Further,
SDs ability to incorporate dynamic changes of key inputs like the demographics
of a patient population makes it useful for high-level healthcare planning over a
long horizon.

3.2 A Simulation Model to Improve Coordination of Health

Services in a Hospital
The previous section described an SD model that considered the potential value of
improving capacity decisions at a long-term, strategic level. In contrast, the
discrete- event simulation model described in this section will focus on some
shorter-term decisions that are more operational in nature. It will focus on a care
pathway for patients who are admitted to the Hospital Medicine Service (staffed by
Hospitalists who provide delivery of comprehensive medical care to hospitalized
patients). This service typically handles patients with multiple comorbidities or
with complex diagnoses that do not fit well into another specialized service.
Therefore, Hospitalist patients were frequently elderly patients or those that
require significant care services upon discharge from the hospital.
The primary purpose of this model is to show how the downstream operations
of an admitting service affect patient flow in the ED. In this model we consider
some of the specific mechanisms causing the boarding delay of ED patients in the
Hospitalist Service at a Calgary hospital. Thus, we are looking at coordinating the
decision making of healthcare services at the process level. Using the model, our
major objectives were:
1. To show the key factors in the Hospitalist admitting service that cause boarding
delays in the ED.
2. Identify opportunities to improve patient flow for both the ED and the
Hospitalist
Service.
As the system model demonstrated in Sect. 3.1, there are interlinkages among
many health services that contribute to delays and impediments to smooth patient
flow. The model in this section considers one of the links in the system in
more detail, so as to increase understanding of these interactions. This will help
understand how the system structure can affect performance at the operational
level and vice versa. As is common in the OR discipline, our objective was to
find improvement solutions that worked well within the whole system and
benefited all of the services involved.
The high-level patient flow for the Hospitalist Service is shown in Fig. 16.10.
Some of the key highlights of the model are:
The targeted capacity of the Hospitalist Service was 180 patients. However, up
to 203 patients could be handled depending on circumstances.
Patients under consideration for admission from the ED could be diverted to
other services when the target capacity limit was reached or passed. From a
patient level of 180203, there was a specified probability that patients would
not be admitted. This probability increased with the number of patients. Once
the hard limit of 203 patients under care was reached, admissions via the ED
and transfers from other services were not accepted. The exception to this was
ICU patients who were always accepted, given the importance of maintaining
ICU availability.

Patients Under Care of

Hospitalists

ED Patients

Discharges

Patients in ED
Admitted to
Hospitalists

Patients In ICU
Transferable to
Hospitalists

Patients in Other
Services
Transferable to
Hospitalists

Patients on
Acute List Under
Care of
Attending
Physician

Patients on Sub
Acute List Under
Care of
Attending
Physician

Alternative Level
of Care
Discharges

Fig. 16.10 Diagram of high-level patient flows within Hospitalist Service

For this reason, ICU patients are given priority when multiple patients are
waiting for a bed.
Two classifications of patients were considered: acute and subacute. Acute
patients required greater care and attention than those considered subacute. A
higher level of staff were required to handle acute patients. Subacute patients
were often waiting to get into an alternative level of care (ALC) space such as a
nursing home bed.
Figures 16.11 and 16.12 show a couple of the key data elements incorporated
in the simulation model of the Hospitalist Service. Figure 16.11 shows the hourly
patterns for the ED and OS (other, non-ED hospitalized) patients that are admitted
or transferred to the Hospitalist Service. The figure shows that ED patients tend
to be spread more evenly across all hours of the day with several small peaks. On
the other hand, the transfers from other services tend to be concentrated during the
hours of 9 am to 6 pm. ICU patient transfers are also concentrated during standard
working hours.
Figure 16.12 shows that patient volumes also varied by day of the week. The
figure shows the pattern for each of the three incoming patient streams to the
Hospitalists. Note the peaks on Fridays for the transfer of patients form other
services (OS) in the hospital. This was, in part, due to a desire for these services
to pass their patients on to another service before the weekend.
Along with the patient arrival data, we incorporated the utilization level of
various hospital resources into the simulation model, in particular, LOS in hospital
beds. Table 16.2 shows the average LOS for patients coming from the ED, ICU,
and other medical services as well as for patients who are eventually classified
as subacute within the Hospitalist Service. Various mathematical functions were

Hourly ED Admission and OS Transfers

0.12

% Per Hour

0.1
0.08
0.06
0.04
0.02
0

Hour
ED
Other
Services

% of Weekly Arrivals

Fig. 16.11 Admission rates to the Hospitalist Service

Arrivals to Hospitalists by
Weekday

0.4

0.35
0.3

0.25
0.2

0.15
0.1
0.05
0
Mon

Tue

Wed

Thu
Fri
Weekda
y
ED OS
ICU

Sat

Sun

Fig. 16.12 Weekly pattern of arrivals to the hospital service

Table 16.2 LOS means for
Hospitalist patients (days)

From
Mean LOS

Acute
ED

ICU

OS

Subacute
All

11.13

13.34

16.27

31.82

evaluated to determine how well they fit the data. The gamma distribution was
found to be a good fit because the LOS data all had a significant right skew and
was therefore used in the simulation model.

Discharges from Hospitalist Service by Weekday

25
%
20%
15%
10%
5%
0%

Sun

Mon
Tue
Hospitalist
Discharges

Wed

Thu
Fri
Sub Acute
Discharges

Sat

Fig. 16.13 Hospitalist discharge pattern

Table 16.3 Discharge patterns evaluated
Discharge pattern

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

Current (%)
Smoothed (%)

14.8
14.3

18.2
14.3

18.1
14.3

17.1
14.3

19.3
14.3

6.8
14.3

5.6
14.3

Finally, the discharge process also had to be built into the model. Figure 16.13
shows the empirical discharge patterns for acute and subacute patients. The
process for modeling the discharge patterns is fairly complex, but requires adding
additional time onto the raw LOS obtained from the data. To avoid biasing the
model with longer lengths of stays than in reality, we adjusted the means and
variance levels of the treatment LOS so that when the discharge adjustment was
made, the overall LOS times in the model would be approximately the same as in
the real system. We tested the discharge modeling approach and were able to
achieve the desired discharge patterns and average LOS times.
A key focus in looking at the decision making for the Hospitalist Service is how
it affected the waiting time for patients in the ED who were eventually admitted to
the Hospitalist Service. We evaluated three options to address these integrated
effects: smoothing discharges, smoothing transfers from other services, and
reducing the LOS for subacute patients.
Table 16.3 shows the current and proposed average discharge rates we tested.
The hourly patterns were not adjusted since we viewed these as more difficult to
change given the various parties that need to come together to complete a
discharge.
Table 16.4 reports the results for the alternative discharge patterns. Smoothing
the discharges more evenly across the week makes a significant difference,
reducing boarding time by over a half an hour on average. Extreme delays are
reduced even more, with the smoother pattern resulting in a 17 h reduction.
Significant reductions in the number boarding and the % of ED patients who are
diverted also result from a more even discharge pattern.

Table 16.4 Results for ED waiting from smoothing discharge pattern in Hospitalist Service
Avg. boarding
Max boarding Avg. #
% ED patients
time (h)
time (days)
diverted
boarding
Scenario
Base
6.95
4.75
3.14
2.21
Smoothed discharge 6.43 0.2a (7.5%)b 4.04 (15.0%)b 2.89 0.1a (8.0%)b 2.00 (9.5%)b
a
b

95% confidence interval halfwidth

Percentage improvement in parentheses

The smoothed discharges across all days are not necessarily optimalit is
likely we could identify an even better weekday pattern of discharges, particularly
given the uneven arrivals to the Hospitalist Service. However, the constant average
values across the weekdays were viewed as a reasonably implementable target.
Figure 16.12 showed that Fridays have significantly more transfers from other
services than other days (nearly one third on Friday alone). However, all things
being equal, smoothing these transfers did not have a positive effect on
performance. Because transfers are somewhat controllable, the current pattern
already adapts to the flow of patients from other sources.
Finally, the results of interviews with Hospitalist Service staff led us to believe
that one of the major causes of delays and poor patient flow from the ED through
the Hospitalist Service was the inability to move patients from subacute care to
ALC options. The primary impact of such a situation in the system would be
lengthy stays in subacute care. This was confirmed by the data once we combined
all stays after a patient was initially classified as subacute. The average LOS
was nearly
33 days for subacute patients. Thus, our final analysis shows the effects of
reducing subacute LOS.
Figure 16.14 shows the results of reducing the average subacute LOS by 5%
increments down to 50% of the current (or base) value (or about 16 days). The
combined effect of discharge smoothing is also shown in the figure. The results
show that ED boarding time is significantly reduced for patients admitted to the
Hospitalist Service as the subacute LOS is reduced. Even a relatively modest 10%
reduction in LOS leads to over a 1.5 h reduction in boarding time for patients
waiting to get into Hospitalist Service beds. If subacute LOS could be reduced by
50%, the impact on the flow of ED patients is dramatic. Boarding times are
projected to be reduced to less than an hour, and on average, nearly two bed
spaces in the ED are freed compared to the base case results.
It is interesting to note that the discharge smoothing essentially has the constant
effect of reducing boarding times by an average of about a half hour regardless of
subacute LOS. Therefore, since there is little or no interaction effect between the
two improvement options, it appears they could be implemented independently
and both significantly improve ED boarding time performance.
In summary, this case used discrete-event simulation to show the value of
coordinating health services for a specific patient pathway in a hospital. This level
of coordination is just as important to creating high-value outcomes as the
previous strategic level coordination model described in Sect. 3.1. While strategic
capacity

Hours of ED Boarding Time

ED Boarding Time vs. Subacute LoS

Reduction

7
6
5
4
3
2
1
0

Base 0.05

0.1

0.15 0.2 0.25 0.3 0.35

0.4 0.45 0.5
Reduction in Subacute
LoS
Base Discharge Smoothed Discharge

Fig. 16.14 Effect of reducing subacute LOS

decisions are certainly constraints for tactical and operating level decisions, if
effective operations management that coordinates the higher level capacity is not
applied at these lower level decision tiers, system performance will be suboptimal.

4 Future Research Challenges

Some emerging trends in healthcare delivery have significant implications for the
coordination of health services. In this section we will discuss two of these trends:
the patient-centered medical home and the centralization of health services at
medical centers.
An emerging movement in healthcare coordination is the concept of a PCMH,
a term first coined by the American Academy of Pediatrics in 1967. Proponents
of PCMH in healthcare share many of the same objectives as health systems
engineers/operations managers, that is, to improve the efficiency and effectiveness
of delivering medical care and, by extension, the health and wellness of the whole
patient.
A PCMH is a team-based model of care led by a personal physician who
provides continuous and coordinated care throughout a patients lifetime to
maximize health outcomes (American College of Physicians 2011). Further, a
PCMH is where the responsibility for coordinating all of their healthcare is
intentionally focused, and serves as the repository for a patients entire medical
record (Hughes and Stiles 1977). A PCMH is, essentially, part of an
engineered health system that incrementally reorganizes and leverages existing
resources to serve the patient.
One area where the PCMH concept will require OM expertise is the design and
management of large groups of patients with chronic conditions. Appropriate
teams

and resourcing will be need to be structured in a coordinated and systems focused

manner. Instead of small groups of primary care doctors caring for chronically ill
patients, teams that include doctors and care managers will likely be used. Homer
et al. (2004) discuss the use of an SD model to help in the transformation in
chronic care services for diabetes and cardiac patients at the county level in the
USA. They highlight that new, focused teams of resources can provide better care
at lower cost; however, it requires a reconfiguration of the healthcare system away
from traditional primary care practice. OM may help in making appropriate tradeoffs among health quality performance and cost to improve the level of healthcare
value.
Another emerging trend in healthcare is for greater centralization of healthcare
services. Major medical centers offer the opportunity to provide inpatient and
outpatient services together using the same facilities (or campus) under a single organizational structure. This allows for unique advantages to care, most importantly
the ability to quickly bring together a multidisciplinary group of physicians and
services to diagnose and treat patients. One such center, Mayo Clinic, was founded
on the concept of integrated health services by housing physicians of all
disciplines in concentrated locations at their three campuses (Jacksonville FL,
Rochester MN, and Scottsdale AZ). This allows for coordination in treatment and
use of resources in ways that standard healthcare systems cannot achieve.
Mayo Clinic and similar centers are considered destination medical because patients will travel significant distances, including internationally, to receive
treatment. High-quality health services are required to attract patients from long
distances; however, operational performance must also be excellent. In particular
receipt of required health services in a short period of time is of prime importance.
This allows patients to plan stays of reasonable length for travel and expense
purposes.
While destination medical centers like Mayo Clinic have many potential advantages, they also have operations challenges related to coordination. To avoid many
patients being scheduled or seeking to use the same services simultaneously,
careful management of physician schedules and capacities is required. To this
end, Mayo Clinic uses advanced decision support systems to help with scheduling
for certain key diagnostic services. Nonetheless, bottlenecks still occur and may
cause delays for patients. To alleviate such circumstances, visits of patients of
different types may be coordinated. For example, national or international
patients coming to Mayo Clinic with similar schedules may cause peaks of
demand for particular services. Local patients can be scheduled to smooth out
these peaks since they have more flexibility with respect to travel.
Another issue in managing medical centers where patients make appointments
well in advance is no-shows, reschedules, and cancellations (NRC). A significant
proportion of appointments end up in these categories and create additional load
on staff to resolve schedules as well as unused capacity due to unfilled
appointment slots. Figure 16.15 shows a causal loop diagram describing the
reinforcing system behavior associated with NRCs. As patient volumes increase,
appointment availabil- ity decreases. Less availability causes delays in finding
appointments (increasing Appointment Latency) which causes more NRCs. This
type of behavior is also

Proportion of NRC
Appointments

+
Long
Wait

Patient Volume

Appointment
Availability

+
Appointment
Latency

Specificity of
Appointment Types

Length of Appointment
Calendars

Fig. 16.15 Causal loop diagram of no-show, reschedules, and cancelation behavior

described in Gallucci et al. (2005). The NRCs take up false capacity and hence
serve to further reduce appointment availability.
This negative cycle may be alleviated by shortening the length of time that
appointments can be scheduled and by creating more flexibility in physician
calendars. However, patients who are traveling from a long distance often want to
make travel plans well in advance of their visit, and physicians prefer to specialize
in treating particular types of patients for research and quality purposes. Thus,
while Open Access scheduling, where little or no time occurs between the
demand and delivery of the health service (see Kopach et al. 2007 for a
discussion), may alleviate NRCs, it may not be easily implementable at large
medical centers. Using advanced scheduling approaches like that discussed in
Helm et al. (2011) that consider the interaction of hospital resources may be more
appropriate for advanced scheduling of patients moving through several healthcare
services.
In summary, emerging trends in healthcare like the PCMH and centralization of
healthcare services into medical centers create new challenges for healthcare
opera- tions management. Operations management, systems engineering, and
systems level modeling can play important roles in how effectively these new
delivery approaches create healthcare value to patients.

5 Conclusions
In this chapter we focused on the value of coordinating healthcare services. With
costs of healthcare rapidly increasing and changes in demographics and policies
that are expanding patient populations, the value of both high-quality and efficient
healthcare systems is paramount. The discipline of operations research will play
a key role in the success of healthcare service integration because of its focus on

coordinating resources to achieve desired outcomes. In the past, this coordination

tended to focus on individual health services. For the future, the coordination of
several services or a system of services will become increasingly important.
It is clear that operations management can play a valuable role in improving the
coordination of existing healthcare services. However, it is likely that new
healthcare systems will evolve in the future, often pushed by the need for radical
change due to increasing costs and policy changes. Our section on research
challenges provided two examples of emerging trends in healthcare delivery
important to operations management. With the patient-centered medical home,
new team structures and care management approaches will be necessary to
success. Another trend toward the growing role of medical centers that patients
travel to creates unique opportunities for integrated care, but significant
challenges in ensuring timely and efficient delivery.
Thus, operations management has much to offer health services in managing
current healthcare process and designing new ones to maximize the delivery of
value to patients.
Acknowledgements The authors wish to thank the following staff from the Calgary Health Region: Karen Abbott, Rob Abernethy, Rick Anderson, Maria Bacchus, Simon Chi, Janet Cohen,
Gil Curry, Stafford Dean, Jim Eisner, Ward Flemons, Jennifer Foster, Toni Furber, Gavin
Greenfield, Nancy Guebert, Beth Harris, Caroline Hatcher, Kathy Howe, Lucy Hyndman, Grant
Innes, Peter Jamieson, Edward Kukec, Linda Lutz, Sharleen Luzny, Lester Mercuur, Micheline
Nimmock, Charlotte Parkinson, Judy Pederson, Lucy Reyes, Tom Rich, Karen Ruggles, Tanya
Sakamoto, Jason Scarlett, Lisa Schaffer, Karalee Schideler, Rick Schorn, Liselotte Simeon,
Dongmei Wang, and Chris Wilkes. In addition, we acknowledge the contribution of Serene Chen
who contributed to this work as a summer intern student at Mayo Clinic.

References
American College of Physicians (2011). http://www.acponline.org/advocacy/where we stand/
medical home/. Accessed 18 Apr 2011
Bekker R, de Bruin AM (2010) Time-dependent analysis for refused admissions in clinical wards.
Ann Oper Res 178:4565
Brailsford SC, Lattimer VA et al (2004) Emergency and on-demand health care: modelling a large
complex system. J Oper Res Soc 55:3442
Cheung PT, Wiler JL, Ginde AA (2011) Changes in barriers to primary care and emergency
department utilization. Arch Intern Med 171(15):13971398
Cooke D, Rohleder T et al (2010) A dynamic model of the systemic causes for patient treatment
delays in emergency departments. J Model Manag 5(3):287301
Davis, K, Schoen, C, Stremikis, K (2010) Mirror, Mirror on the Wall: How the Performance of
the U.S. Health Care System Compares Internationally, 2010 Update, Commonwealth Fund
Report, June 2010
Gallucci G, Swartz W, Hackerman F (2005) Impact of the wait for an initial appointment on the
rate of kept appointments at a mental health center. Psychiatr Serv 56:344346
Haraden C, Resar R (2004) Patient flow in hospitals. Frontiers of Health Serv Mgmt 20(4): 315
Helm JE, AhmadBeygi S, Van Oyen M (2011) Design and analysis of hospital admission control
for operational effectiveness. Prod Oper Manag 20(3):359374

Hillestad R, Bigelow J, Bower A, Girosi F, Meili R, Scoville R, Taylor R (2005) Can electronic
medical record systems transform health care? Potential health benefits, savings, and costs.
Health Aff 24(5):11031117
Homer J, Hirsch G et al (2004) Models for collaboration: how system dynamics helped a
community organize cost-effective care for chronic illness. Sys Dyn Rev 20(3):199222
Hughes JR, Stiles F (1977) Fragmentation of care and the medical home. Pediatrics 60:559
Institute of Medicine (1999) To err is human: building a safer health system. National Academy
Press
Kelly A-M, Bryant M, Cox L, Jolley D (2007) Improving emergency department efficiency by
patient streaming to outcomes-based teams. Aust Health Rev 31(1):1621
Kopach R, DeLaurentis P-C, Lawley M, Muthuraman K et al (2007) Effects of clinical characteristics on successful open access scheduling. Health Care Manag Sci 10(2):111
Lane DC, Monefeldt C et al (2000) Looking in the wrong place for health care improvements: a
system dynamics study of an accident and emergency department. J Oper Res Soc 51:518531
Lang M (2006) City emergency care under review: probe follows high-profile problems. Calgary
Herald, Calgary
Levin S, Dittus R et al (2011) Evaluating the effects of increasing surgical volume on emergency
department patient access. BMJ Qual Saf 20(2):146152
Meredith JR,Shafer SM (2007) Operations management for MBAs, 3rd edn. Wiley, New Jersey
Morecroft J (2007) Strategic modelling and business dynamics: a feedback systems approach.
Wiley, West Sussex, England
Sterman JD (2000) Business dynamics: systems thinking and modeling for a complex world.
McGraw-Hill, New York
Rohleder TR, Bischak DP et al (2007) Modeling patient service centers with simulation and
system dynamics. Health Care Manag Sci 10(1):112
Rohleder TR, Rogers P, Cooke DL, Xu S (2009) Emergency department simulation models: a
report for the Calgary Health Region. Report commissioned by the Calgary Health Region
Wolstenholme EJ (1993) A case study in community care using systems thinking. J Oper Res Soc
44:925934